| Abstract: | Subjects solved the hobbits-orcs problem repeatedly until they made no errors on two successive solutions. Groups varied in the kind of feedback given after errors. Further support was obtained for Thomas' (this journal) conclusion that subjects organize the sequence of moves in small subsequences, rather than individual moves. Differences between feedback groups suggested that the organization is oriented forward in the problem, and this contrasts with GPS' retroactive organization of this problem. Comparison between feedback groups also leads to the conclusion that subjects learned from positive information indicating which response is correct, rather than by a process of eliminating errors or sampling new strategies after errors. Analysis of acquisition data agreed with the hypothesis of all-or-none learning at individual states, except for one state where the additional complexity was related to an ambiguity about backward moves. The general pattern of results was invariant over a change in the characters of the problem (men and elves) designed to produce a reversal of relationship between the characters concerning who was prevented from outnumbering whom. |
